Big data analytics methodologies applied at energy management in industrial sector: A case study

Abstract

In this work, a framework is developed to integrate IoT-based energy management and company’s existing information systems. This framework is a multi-layer model that includes three layers: 1) data collection layer, 2) data management layer and 3) data analytics layer. In order to test the proposed approach and assess its impact on improving energy efficiency, a pilot study was carried out in an Italian manufacturing company. Several smart meters have been installed at machine level to collect energy consumption data in real time, and then this data have been analyzed and provided to decision makers to improve energy efficiency by integrating them in production management decisions. When a company aims at analyzing the energy characteristics of its production system, data provided by different sources and geographically dispersed repositories must be taken into consideration. These characteristics bring several problems to develop a data analytic architecture. In this paper, we propose a data analytic model for IoT, in order to integrate the data collected from different sources and to improve energy-aware decision-making. Improving the overall equipment effectiveness of machine tools will improve resource-efficiency and productivity in manufacturing and support the development of smart factories from an energy point of view.

Keywords

Internet of things data analytics data mining energy management big data management industrial energy consumption

1 Introduction

Worldwide about 50% of the total electricity consumption is made in industry by conversion using electric motor-driven systems of workstations (Waide & Brunner, 2011). There have been significant efforts over the last decade to define appropriate standards and best practices and implement consistent energy management systems to increase and maintain energy savings. Company energy management systems seek approaches to reduce their energy consumption without declining the production outputs (Hadera et al., 2015). According to Vikhorev et al. (2013), companies should define an energy management framework to promote energy awareness in manufacturing processes.

In this context, this work aims at developing a framework to improve the energy efficiency of a production system using concepts of Internet of Thinks (IoT) and Data Analytics. Relationships between IoT and Data Analytics is described by Feller et al. (2015): IoT collects data from different sources (in this work from power and energy sensors embedded in workstations), Data Analytics is responsible for extracting patterns or generating models from the output of the data processing step and then feeding them into the decision-making step, which takes care of transforming its input into useful knowledge.

The first step towards reducing energy consumption in machine tools and manufacturing systems is to devise methods to understand and characterize their energy consumption (Herrman et al., 2007). For this reason, an energy monitoring system is necessary. IoT can provide useful tools in order to develop a more detailed analysis of machine consumption. Kees et al. (2015) define IoT as “the connectivity of physical objects or industrial products, equipped with sensors and actuators, to the Internet via data communication technology, enabling interaction with and/or among these objects” or products. Moreover, electrical energy metering in complex manufacturing facilities is necessary to provide industrial enterprises higher levels of quantification and visibility in their energy consumption. Both voltage and current need to be measured at either low or high sampling rates, in order to calculate power consumption and to produce more complex power quality statistics such as sags, peaks, and harmonics (O’Driscoll & O’Donnell, 2013). On the basis of the measured power, empirical energy models can be built for estimating the energy consumption related to the production.

There has been a compelling need to adopt data management systems in industrial operational processes and product-development principles in order to enhance IoT applications, while the development of big data is already lagging behind in integration with cloud computing. It has been widely recognized that these two technologies are interdependent and should be jointly developed: meaning that the widespread deployment of IoT drives the high growth of data (Min et al. 2014). Big data management is a complex process, particularly when abundant data originating from heterogeneous sources are to be used for business intelligence and decision-making (Baker, 2014). Furthermore, big data management has become a key to the success of many enterprises, science, industries, engineering fields and government ventures (Chaudhuri et al., 2011). The main objective is to enhance data quality and accessibility for decision-making and improve productivity. Therefore, “big data” could be defined as a fast-growing amount of data from various sources that increasingly poses a challenge to industrial organizations and also presents them with a complex range of valuable-use, storage and analysis issues.

Traditional data storage and processing are typically fed with relatively clean datasets generated by limited sources; hence, the results tend to be accurate. However, the recent introduction of Industry 4.0 paradigm leads to the collection of massive, heterogeneous and frequently generated data (Bevilacqua et al., 2017). This has revealed a serious management problem not only due to the growth in the volumes of datasets but also to their complexity and volatility that makes processing and analysis very hard to achieve. These aspects are very important when a company aims at analyzing the energy characteristics of its production system. In this case, data provided by different sources and geographically dispersed repositories must be taken into consideration. Important information must be collected from administrative office (i.e. electricity accounts), production sites (i.e. workstations energy consumption, marking data of production progress), from suppliers (i.e. material delivery date), from production planning department (i.e. master production schedule) and from technical department (i.e. products codes, bill of materials, working times and manufacturing cycle).

These characteristics bring several problems to develop data analytics architecture. In this paper, we propose a data analytics framework for IoT, in order to integrate the data collected from different sources and to improve energy-aware decision-making.

The remainder of this paper is organized as follows. Section 2 presents a literature review regarding energy consumption models developed for production systems, with a special focus on IoT, Big Data Management and energy management in the industrial sector in Section 2.1. Section 3 introduces the research approach and the case study. Results of the study are shown in section 4 while discussion and conclusions are reported in Section 5.

2 Energy consumption models for production systems: Literature review

The development of energy models at the level of unit process is a research topic quite analyzed in literature.

Some authors focused on machining process level while other works were carried out on the machine-tool level. Regarding papers on machining process level, Srinivasan and Sheng (1999) developed an approach for macro and microplanning of feature-based machining. Microplanning looked at selecting process parameters, tooling, and cutting fluid based on process energy use, waste streams, process quality, and machining time. Draganescu et al. (2003) used experimental data and Response Surface Methodology (RSM) to establish a statistic model of machine-process efficiency and specific energy consumption in machining. Sarwar et al. (2009) chose Specific Cutting Energy (SCE) as the evaluation parameter of measuring the efficiency of the metal cutting process, and the variation of SCE regarding different workpiece materials provided valuable information for bandsaw manufacturers and end users to estimate machinability characteristics for selected workpieces.

Regarding works on machine-tool level, we can highlight Diaz et al. (2011) work. They carried out a characterization on the energy consumption of milling machine tools during their use stage. They studied the effect of workpiece material on power demand. Dahmus and Gutowski (2004) developed an experimental research on machine tool energy consumption and categorized the total energy of the system in three main activities, namely “Constant start-up operations”, “Run-time operations” and “Material removal operations”. Herrmann et al. (2010) addressed the energetic consumption of the machine tools and extended the perspective by considering the ecological aspects beyond the economic input and output flows.

Assessment methods of energy consumption are another aspect quite analyzed in literature. Some authors highlighted the importance of carrying out a real-time monitoring system while other authors proposed theoretical models for analyzing energy consumptions.

For instance, in Abele et al. (2012), work power measurements of a single machine tool is described by several functional modules that further consist of various components. Within their Hardware-in-the-Loop-Simulation (HiLSimulation), a physical machine controller is connected to the simulation model so that the programmable logic control (PLC) or numerical control (NC) signals, which contain power-on states, axis speeds, machine tool movement path, process operations, etc., are coupled with the functional modules and components to enable continuous energy simulation of a machine tool. In addition to estimating the machine energy requirement within the work of Abele et al. described above, Eberspacher et al. (2014) further developed the HiL-Simulation model for real-time monitoring of the energy demand of a machine and its functional modules in production environments. Dietmair and Verl (2009) introduced a generic method to model the energy consumption behavior of machines based on a statistical discrete event formulation. The parameter information required to characterize the discrete events can be obtained with a small number of simple measurements or with a degree of uncertainty from the machine and component documentation. Vijayaraghavan and Dornfeld (2010) pointed out that, in order to decrease energy consumption, energy data has to be placed in the context of the manufacturing activity. They developed a monitoring system, in which MTConnect_standard, as an XML-based standard, for data exchange is selected for data collection from manufacturing equipment.

The automated monitoring system can help attach contextual processing-related information to the raw data. Therefore, it is very important for reducing energy consumption in order to develop a real-time energy efficiency monitoring system of machine tools. On the other hand, several authors highlighted that generally manufacturing machines and equipment are not metered permanently (Müller & Loffler, 2009; Garetti and Taisch (2012). Lack of sub-metering is highlighted as the main barrier to improving energy efficiency in non-energy-intensive manufacturing by Rohdin and Thollander (2006), as well for energy intensive industry by Trianni et al. (2013).

2.1 IoT and data analtics for energy management in production systems

Recently, some authors tried to integrate IoT and Big Data Management concepts in order to manage energy consumption data and improve energy-aware decision-making. Tao et al. (2014) developed a new method for Energy-saving and emission-reduction based on Internet of Things (IoT) and bill of material (BOM). Shrouf et al. (2014) proposed an approach for energy management in smart factories based on the IoT paradigm. They developed a guideline and highlighted expected benefits of this approach.

Always Shrouf and Miragliotta (2015) contributed to the understanding of energy-efficient production management practices that are enhanced and enabled by the Internet of Things technology. Moreover, they presented a framework to support the integration of gathered energy data into a company’s information technology tools and platforms.

Regarding Data Management in literature data analytics methods have been used by various researchers in energy system applications. With the emergence of the data mining approach for predictive modeling, different types of models can be built on a unified platform: to implement various modeling techniques, assess the performance of different models and select the most appropriate model for future prediction. Tso and Yau (2007) have used regression analysis, decision tree and neural networks models for the prediction of electricity energy consumption. Model selection is based on the square root of the average squared error. Figueiredo et al. (2005) presented an electricity consumer characterization framework based on a knowledge discovery in databases procedure, supported by data mining techniques. Lu et al. (2013) presented a framework for predicting the electricity price, the price spike, the level of spike and the associated forecast confidence level. The proposed model is based on a mining database including market clearing price, trading hour, electricity demand, electricity supply and reserve.

Other efforts (Seem, 2007; Berglund et al., 2011) first extract the features from daily energy consumption then use statistical methods to identify abnormally high or low energy use. However, these methods relied on the assumption that the data is sampled from a particular distribution, which may not hold true.

3 Energy management framework

Literature in the energy management research field focused attention on developing methods for reducing energy consumption and improving energy-aware decision-making. No many works developed integration methods of complex data sets from multiple information sources such as energy system, production system and enterprise information systems. These data sets must be integrated with data streams collected from wired and wireless sensors and meters in order to perform N-dimensional analysis of energy performance data and to support the decision-making process of the end users. Moreover, modern energy management systems incorporate data archival but energy managers need assistance in extracting useful information from a large volume of data compiled. In this context, this paper addresses this deficit by introducing a new method for designing a Data Warehouse that can be useful to a production system that aims at collecting big data. In our case study, data comes from many sources affected by veracity problems and are provided with a different velocity. The energy management framework proposed in this work is based on a data warehouse to store, integrate and analyse the complex data sets in order to support multi-dimensional analysis of energy performance data.

An approach is developed to integrate IoT-based energy management and company’s information technology tools and platforms. This approach has been used in the case study presented in the next section but it is a general framework that can be used in every manufacturing company. The framework (Fig. 1) is a multi-layer model that includes three layers: 1) data collection layer, 2) data management layer and 3) data analytics layer. According to Haller et al. (2008), Internet of Things model has generally been recognized as three layers. The bottom level is used to perceive sensory data; the second layer is the network layer for data transmission; the top is the application layer. In this perspective, Data Collection layer implements the first two layers of the Haller architecture, while Data Management and Data Analytics layers belong to the Haller’s application layer.

Fig.1

Muti-layer model.

3.1 Data collection layer

Data collection layer adopts company existing information systems and devices, e.g. RFID Reader (Manufacturing Execution System) and sensors (Workstations), to collect various smart object’s data. With the fast development and pervasive application of information technology in manufacturing, enterprise information systems such as product data management (PDM), enterprise resource planning (ERP), computer-aided design (CAD), computer aided process planning (CAPP), customer relationship management (CRM), and supply chain management (SCM) have been widely accepted and applied by manufacturing enterprises. Different types of data require different data collection strategies. Moreover, designing the layer require to define: (1) the machines that will be monitored; (2) a list of required measures (active power, reactive power, etc.); (3) the monitoring devices for each machine and its specifications; (4) the communication system; (5) where and how the data will be stored and analyzed. Furthermore, the production processes have to be identified (e.g. production sequence, the processing time for each product under different machine configuration), so as to link and understand the energy consumption behavior and make the efficient decision. In the case study proposed in next section, it is assumed that each individual sensor will be assigned its own unique IP address and that the data will be transported through the Internet infrastructure to cloud-based applications. One of the pillars of IoT is cloud computing. According to Cook and Das (2004), cloud computing platforms provide two types of benefits: managed infrastructure services and a software framework that simplifies the development of large-scale applications. Cloud computing builds on the virtualization capabilities of modern computer systems to provide organizations with on-demand computing and storage capabilities. Through the use of fault tolerant systems and geographically distributed data centers, it can ensure high availability. Most importantly, it ensures that updates and patches can be applied in a timely manner.

Companies may benefit from cloud-based solutions without compromising security using a multi-layered approach and private clouds. They may choose to manage their own internally hosted data acquisition applications to collect data from their own private network of devices and use some type of gateway application to push the data to a cloud-based application for processing.

3.2 Data management layer

In the process of data collection, a series of problems, e.g., energy-efficiency, misreading, repeated reading, fault tolerance, data filtering and communications etc., must be solved. Data management layer applies an ETL (Extraction, Transformation and Loading) process for pulling data out of the source systems and placing it into a Data Warehouse (DW). After the extraction of interesting data from sources, they are cleaned and transformed for reconciling possible semantic heterogeneities (e.g., synonyms and homonyms, different representation of semantically equivalent concepts). Finally, data are saved in the DW. For instance, RFID data are collected by sensors as a stream of EPC format, which is the universal identifier for a physical object (Bevilacqua et al., 2013). After transformation, data are structured in a table where each record contains EPC, location, time_in and time_out. A DW is a system used in the enterprise to support decision at strategic level, hence it provides a unique view of the entire organization, including also external data. Furthermore, DW provides the access to current and historical data, which are typically stored in aggregated form. Data stored in a DW are used for creating analytical reports for knowledge workers throughout the enterprise. Data in a DW are organized on the basis of the multidimensional model (Golfarelli et al., 1998).

3.3 Data analytics layer

Data analytics layer is built based on data management and event processing. Various object-based or event-based data mining services, such as classification, forecasting, clustering, outlier detection, association analysis or pattern mining, are provided for applications, e.g., supply chain management, inventory management and optimization etc. The architecture of this layer is service-oriented. Smart meters and sensors enable remote monitoring of energy consumption data across the factory. The data can then be stored and analyzed. The results and warning messages can be delivered through mobile applications to shop floor supervisor. Also, energy management experts can make real-time assessment by having a clear picture of energy consumption in real time.

After collecting and analyzing data, in this phase data are exploited into energy management tools (e.g. energy decision support system, simulation tools) to enable the decision makers to enlighten possible waste of energy, where improvement can be achieved, or select the most sustainable configuration mode of machines by considering the production planning in order to improve energy efficiency. In this phase, decision makers can also define strategies and practices to improve the energy efficiency of the smart factory “by design”, for example by integrating energy data in production management practices.

4 Case study

A case study was conducted to show the proposed method. The company analyzed is an Italian medium enterprise specialized in the production of turned metal parts and precision mechanical components.

The production system of the company is strongly oriented to high technology products thanks to the presence of numerical control single and multiple spindle machines. The company performs automatic CNC turning operations on different types of metals: Aluminum, Copper, Brass, steel ETG 100, quenched and tempered steel, Stainless Steel, Alloy Steel. The company buildings cover an area of 5548 m² and have an electrical total annual consumption equal to 2,094,936 Mwh. Figure 2 shows the layout of the production site. The company is made up of four operating areas: raw material warehouse, processing departments, finishing and washing department, finished products warehouse; represented by spots 2, 3, 4 and 5 in Fig. 2. Company departments work 230 days per year in three shifts of eight hours/day. Machines set-up is carried out exclusively during the morning shift.

Fig.2

Company layout.

Various sensors have been installed on production machines since 2015 to acquire electrical measurements through single-phase and three-phase multimeters. The position of the sensors is shown by the spots in Fig. 2. Low-cost sensors have been used in order reduce the payback period of the investment. The meters are connected to a gateway that links the Modbus network with Ethernet. A web platform has been designed for acquisition, processing and presentation of energy and production data according to Big Data paradigms. Indeed, the case study is a typical Vs-problem. The velocity and volume is due to the number of acquisitions; in addition to the 10 meters shown in the Figure, we have one sensor for each active machine. Each sensor collects three measures: active power, reactive power and energy consumptions. Although in the case study the sensors sampling rate is low, we like to note that, the proposed framework is independent on the velocity of acquisition. Furthermore, as shown in next subsection, we take into account different forms of data (variety): energy sensors’ measures, production data from the ERP and manually inserted production phases.

4.1 Data sources

In the case study, both information from the ERP and data from sensors has been considered. From the former, the production plan has been extracted in order to be aware of what has been produced and characteristics of the product, like name, description and raw material. Sensors have been used to collect actual data: production events and energy consumption.

A production event is characterized by an operator, which manages a specific production phase, in a given time interval. Data about events has been collected using several QR codes placed next to each machine; a QR code for both the beginning and the end of each specific phase. In this case study, four phases have been identified: set-up, warming-up, production and technical stoppage. Furthermore, stoppages are further classified in sub-phases, namely machine tuning, changing tool, production completed, separator replacement, plank manual replacement, and loading jam. The operator, by scanning the QR code through a common reader, easily identifies the phase or sub-phase and the system assigns the actual timestamp, so all information about the event are collected.

Data about energy consumptions are collected by means of sensors which are able to measure active power, reactive power and energy consumptions. Nine machines have been coupled with an energy sensor. In order to ensure the economic sustainability of the experiment, while ensuring the quality, the company has chosen to adopt sensors that return accurate measurements but with low frequencies; the sampling time is 15 minutes. Chosen sensors return the cumulative values in the sampling interval of active power, reactive power and energy consumption.

All data from sensors are sent to and stored in an enterprise data cloud. Figure 3 shows the conceptualization made for data in the cloud using the Entity-Relationship (ER) model, a reference model for conceptual design (Chen, 1976). Similarly, Fig. 4 shows the portion of the ERP schema representing needed information about the production plan. From Fig. 3(a), a production event is identified by the triple formed by code of the machine, start and end date/time, and is characterized by the code of the article that is being produced, the related order and the phase to which the event refers. As regards energy consumption in Fig. 3(b), each sensor value is identified by the related machine and the instant the measurement refers to; each measurement is formed by three values: energy, active power and reactive power.

Fig.3

ER representation of sensor data schema: (a) production event, (b) energy consumption.

Fig.4

ER representation of the portion of the ERP schema representing the production plan.

In Fig. 4, a production order describes the number of products (“quantity”) of a given article that will be produced on a specific machine on the scheduled date (“SchedDate”). A production order is identified by its number (“OrdNum”). Each article is characterized by a textual description, the code, and the raw material of which is made the cylinder used in the lathe and its diameter. Finally, a production order is assigned to a given machine, which is located in just one department.

4.2 Data Warehouse design

The first step of Data Warehouse design (Golfarelli et al., 1998; Kimball & Ross, 2002) is the integration of relevant portions of different data schemas. Conceptualization in the form of ER schemas greatly simplify this step, allowing to easily identify overlapping elements, conflicts in the representation of the same element, and/or missed relationships between different schema fragments. For instance, it can be seen that a machine is represented as a separate entity with its own attributes in ERP schema (Fig. 4) while it is a simple attribute describing the name of the machine where a Production Event takes place (Fig. 3(a)). The general strategy is to choose, among the different conflicting representations, the most general one, that is the one that allows accommodating the others as special cases (in the example, the machine as an entity). Similar reasoning applies to other conflicting representation. A more critical issue is the integration of Production Events (Fig. 3(a)) with their energy-related measures (Sensor Value in Fig. 3(b)). Indeed, events refers to highly-variable (manually defined) time intervals, while measurements are acquired every 15 minutes from 00 : 00 of each day. Figure 5 shows an example of production event and measurement interval (i.e., the time between two subsequent measurement instants). In the Figure, the start and end time of a production event (i.e., a production interval) is represented using subscript c, while the subscript p is used for measurements. In order to integrate data, we have intersected intervals hence defining new events which inherit the production phase from the corresponding production event and consumption values from measurement intervals. In particular, energy (active power, reactive power) consumption for each interval is computed as a fraction of the measured energy (active power, reactive power) that is proportional to the length of the interval itself. The following formula is used to estimate consumption values of the interval I_i = [t_i; t_i +1]:

Fig.5

An example of integration of consumption (t_ci) and production event (t_pj) timelines.

$(\begin{matrix} \hat{E} (t_{i}) \\ \hat{AP} (t_{i}) \\ \hat{RP} (t_{i}) \end{matrix}) = (\begin{matrix} E (t_{c (x + 1)}) - E (t_{cx}) \\ AP (t_{c (x + 1)}) - AP (t_{cx}) \\ RP (t_{c (x + 1)}) - RP (t_{cx}) \end{matrix}) . \frac{(t_{i + 1} - t_{i})}{15}$ where $\hat{E} (t_{i}), \hat{AP} (t_{i})$ and $\hat{RP} (t_{i})$ are estimated energy, active power and reactive power in the interval I_i respectively, E(t_cx), AP(t_cx) and RP(t_cx) are energy, active power and reactive power measured by a sensor at the x-th instant, and $t_{c (x + 1)} = {k \cdot 15, k \in ℕ | k \geq \frac{t_{i + 1}}{15}}$ . All time values are in minutes. For instance, in Fig. 5, the interval [t₉, t₁₀] inherits the phase of the production event in [t_p3; t_p4] and energy estimate $\hat{E} (t_{9}) = (E (t_{c 7}) - E (t_{c 6})) \cdot \frac{(t_{10} - t_{9})}{15}$ . Noteworthy that a simpler integration strategy, such as assigning to a production event the average of all measurements gathered during its interval (e.g., in [t_p1; t_p2] there are two measurements in t_c2 and t_c3), would lead to a loss of information; indeed, during a fast production event (, like [t_p2; t_p3]), no consumption values could be collected, hence we would not able to assign estimated energy, active power and reactive power to the event. In these cases, we would not be able to correctly perform all analysis (e.g. to evaluate energy consumption with respect to production phases), hence resulting in low-quality DW.

The resulting integrated schema is shown in Fig. 6, where “Article” and “Machine” attributes, of “Production Event” and “Sensor Value” respectively, have been reificated by using the extended descriptions provided by the ERP schema.

Following the approach proposed by Golfarelli et al. (1998), next steps in Data Warehouse design are related to the definition of multidimensional model elements, namely the fact to be analysed, the analysis perspectives (or dimensions), the granularity levels at which data are shown, and the measures by which the fact is evaluated. The pivotal concept in the integrated schema, which is chosen as fact, is the “Event”. Note that, one might think to keep the “Machine” at the center, but since multiple events are related to the same machine, this design choice would lead to a DW where each machine will have several phases and just aggregated consumption values (e.g., average), and the analyze of consumption with respect to the phases will be prevented.

Fig.6

The integrated schema.

Four dimensions have been selected: the machine where the production occurred, the time when the event begins, the product and the production phase. Each dimension is structured in a hierarchy of levels, where a part-of relation exists between members of a level and members of a higher level (e.g., a machine is part of a department). In this way, moving from a level to the higher one, the fact represents a wider portion of data and corresponding measures’ values are aggregation of values returned at lower level. The following hierarchies have been selected:

Machine: department→ machine;

Time: year→ month→ day→ time slot→ starting time;

Product: article→ raw material;

Phase: phase→ sub-phase;

In order to allow analysis at different granularity levels, the “Phase” and “StartingTime” attributes have been suitably expanded. Measures have been selected among quantitative attributes of the fact. In particular:

Total Energy: = sum( $\hat{E}$ )

Total ActivePower: = sum( $\hat{AP}$ )

Total ReactivePower: = sum( $\hat{RP}$ )

Time interval: = sum(δ)

Average Power: = $\frac{Total Energy}{Duration}$

The final star schema of the Data Warehouse is depicted in Fig. 7.

Fig.7

The Data Warehouse star schema.

In order to populate the DW, an ETL process has been defined, which is able (a) to extract useful data from sources (Figs. 3 and 4), (b) to transform them removing errors, defining new time intervals and estimating new energy consumption values (as defined above), (c) to load data into the DW (Fig. 7).

In particular, some errors could occur in source data due to failures of energy consumption sensors and to the manual gathering of data about production events. In the first case, measures detected by the sensor could be missing or out-of-the-range. In the ETL process, they are replaced with values obtained by the same sensor in the previous sampling interval. As regards production data, there are two main issues: 1) the operator wrongly scans the begin of a phase before the end of the previous one; 2) the same system is used to record the end of a work shift, in this case we have a production event which begins and ends at the same time. In the first case, usually the operator quickly recognizes the error and scans the right code; hence, these kind of issues are handled by reversing overlapping timestamps. Issues of the second kind are simply removed, because they do not represent production events.

4.3 Analysis

OLAP analysis (Fig. 8) allows the company to divide the total machine tool power in four power levels:

Fixed power: power demand of all activated machine components ensuring the operational readiness of the machine.

Operational power: power demand to operate components.

Tool tip power: power demand at tool tip to remove the workpiece material.

Unproductive power: power converted to heat mainly due to friction during the material removal.

During the implementation of the proposed approach, several operations management practices have been enhanced by integrating energy data. For example, energy consumption data has been collected from the machines under different configurations (e.g. machine speed). Then, based on the flexibility of production schedule, this data enabled the production manager to select the most efficient configuration of the machines. Moreover, the analysis of energy consumption data provided the decision makers with a clear picture on the energy waste at production level (e.g. idle time), and it provided precise information about the energy needed to produce one piece. Figure 9 illustrates a typical energy consumption profile of a turning machine.

Fig.8

A screenshot of the OLAP analysis implemented using Mondrian.

Fig.9

Energy consumption profile of a turning machine.

The load profiles of single machines add up to a cumulative load profile for the process chain and determine the embodied energy of a product. The specific energy and resource consumption behavior of a process chain is significantly influenced by its specific technical configuration (design) and control. This includes the individual selection/combination of processes/machines and their inter linkage (e.g. process chain structure, buffers) as well as aspects like batch sizes, scheduling of orders (e.g. start time, capacity allocation) or speed of production.

In this perspective, the electrical work can be reduced, for example by optimal utilization of equipment and avoiding energy waste in idling machines or the selection of appropriate machines for the specific manufacturing task. The avoidance of consumption peaks is another important issue. From an economic perspective, peaks should be avoided since they may cause cost surcharges in an electricity bill and production interruptions. Other important aspects concern the possibility to shift the energy consumption from day to night, because of less expensive energy price rates (e.g. base time at night).

5 Discussion and conclusion

The paper presented a general methodology and practical issues involved in the development of a data analytics framework for IoT in an energy management setting.

The results of this study can facilitate various applications for supporting an energy aware design and control of process chains. Energy efficiency improvement requires awareness of energy consumption behavior at production line and machine level. In this context, smart meters provide real time data, and take decisions in collaboration with company information systems. In the case study, combined with other information systems, Internet of things played the role of data aggregation, improving management capacity and efficiency. The core technology is the energy sensor network and computer information processing, for building an advanced, powerful information acquisition and processing platform. New data analytic model has been developed to integrate the data collected from different sources: workstations energy consumption and marking data of production progress from production site, electricity accounts from administrative office, material delivery date from suppliers, master production schedule from production planning department and products codes, bill of materials, working times and manufacturing cycle from technical department. The process commences by transforming big data from original format to computer formats. It progresses with applying big data operations towards achieving decision-making. We propose a big data management process flow as a layered component diagram that shows all steps big data must undergo in order to accomplish the management process. The journey begins with big data being transmitted from different sources to storage devices and continues with the implementation of pre-processing, integration and analysis, amounting to the decision-making endpoint.

There are some advantages of the proposed data analytics based method. First, the machine energy patterns and characteristics can be discovered by this method, while traditional methods can only show the statistical characteristics of the entire data set. Second, different analysis perspectives can be adopted, focusing on products or product types, or the whole production process, besides machines. Finally the approach is flexible enough to allow the introduction of new sensors and/or measures, or new kinds of analyses. We plan to exploit the system deployed to obtain further insights about energy consumption profiles and patterns, also by applying different data analysis techniques like clustering and predictive analytics techniques.

References

Abele

, Eisele

, & Schrems

(2012) Simulation of the energy consumption of machine tools for a specific production task. In: Dornfeld

D.A.

, Linke

B.S.

(Eds.), Leveraging Technology for a Sustainable World. Springer, Berlin Heidelberg, pp. 233–237.

Baker

(2014). Designing and managing Big Data – How are you researching your outcomes. In SimTecT.

Berglund

, Michaloski

, Leong

, Shao

, Riddick

, Arinez

, & Biller

(2011). “Energy Efficiency Analysis for A Casting Production System,” in Winter Simulation Conference, IEEE, pp. 1060–1071. Proceedings of the 2014 IEEE IEEM 701.

Bevilacqua

, Ciarapica

F.E.

, Mazzuto

, & Paciarotti

(2013). The impact of RFID technology in hospital drug management: An economic and qualitative assessment. International Journal of RF Technologies, 4(3-4), 181–208.

Bevilacqua

, Ciarapica

F.E.

, & De Sanctis

(2017). Lean practices implementation and their relationships with operational responsiveness and company performance: An Italian study. International Journal of Production Research, 55(3), 769–794.

Candela

, Castelli

, & Pagano

(2012). Managing big data through hybrid data infrastructures. ERCIM News, 89, 37–38.

Chaudhuri

(2012). What next?: A half dozen data management research goals for big data and the cloud. In: Proceedings of the 31st Symposium on Principles of Database Systems. ACM.

Chaudhuri

, Dayal

, & Narasayya

(2011). An overview of business intelligence technology. Commun ACM, 54(8), 88–98.

Chen

(1976). The entity-relationship model - toward a unified view of data. ACM Transactions on Database Systems, 1(1), 9–36.

10.

Cook

D.J.

, & Das

S.K.

(2004). Smart Environments: Technologies, Protocols, and Applications, in: Wiley Series on Parallel and Distributed Computing, Wiley-Interscience, 2004.

11.

Dahmus

, & Gutowski

(2004). An Environmental Analysis of Machining. Proceedings of ASME International Mechanical Engineering Congress and R&D Exposition, 13–19.

12.

Diaz

, Redelsheimer

, & Dornfeld

(2011). Energy Consumption Charaterization and Reduction Strategies for Milling Machine Tool Use. 18th CIRP LCE Conference, Braunschweig, 263–267.

13.

Dietmair

, & Verl

(2009). A generic energy consumption model for decision making and energy efficiency optimisation in manufacturing. Int J Sustain Eng, 2, 123–133.

14.

Draganescu

, Gheorghe

, & Doicin

C.V.

(2003). Models of machine tool efficiency and specific consumed energy. Journal of Materials Processing Technology, 141(1), 9–15.

15.

Eberspacher

, Schraml

, Schlechtendahl

, Verl

, & Abele

(2014). A model- and signal-based power consumption monitoring concept for energetic optimization of machine tools. Procedia CIRP, 15, 44–49.

16.

Feller

, Ramakrishnan

, & Morin

(2015). Performance and energy efficiency of big data applications in cloud environments: A hadoop case study. Journal of Parallel and Distributed Computing, 79–80, 80–89.

17.

Figueiredo

, Rodrigues

, & Gouveia

J.G.

(2005). An Electric Energy Consumer Characterization Framework Based on Data Mining Techniques. IEEE Transactions on Power Systems, 20(2005), 596–602.

18.

Garetti

, & Taisch

(2012). Sustainable manufacturing: Trends and research challenges. Prod Plan Control Manag Oper, 83–104.

19.

Golfarelli

, Maio

, & Rizzi

(1998). The dimensional fact model: A conceptual model for data warehouses. International Journal of Cooperative Information Systems, 7(2-3), 215–247.

20.

Hadera

, Harjunkoski

, Sand

, Grossmann

I.E.

, & Engell

(2015). Optimization of steel production scheduling with complex time-sensitive electricity cost. Comput Chem Eng, 76, 117–136.

21.

Haller

, Karnouskos

, & Schroth

(2008). The Internet of Things in an Enterprise Context,” in First Future Internet Symposium, FIS 2008 Vienna, Austria, September 29-30, 2008 Revised Selected Papers, Domingue

, Fensel

, and Traverso

, Eds. Springer-Verlag, Berlin, Heidelberg, 2009, 2009, 14–28.

22.

Herrman

, Bergmann

, Thiede

, & Zein

(2007). Energy Labels for Production Machines – An Approach to Facilitate Energy Efficiency in Production Systems. Proceedings of 40th CIRP International Seminar on Manufacturing Systems Location, Liverpool, UK.

23.

Herrmann

, & Thiede

(2009). Process chain simulation to foster energy efficiency in manufacturing. CIRP Journal of Manufacturing Science and Technology, 1(4), 221–229.

24.

Herrmann

, Kara

, Thiede

, & Luger

(2010). Energy Efficiency in Manufacturing Perspectives from Germany and Australia. 17th CIRP LCE Conference, Hefei, 23–28.

25.

Kees

, Oberländer

, Röglinger

, & Rosemann

(2015). Understanding the Internet of Things: A Conceptualisation of Business-To-Thing (B2T) Interactions. In the Proceedings of Twenty-Third European Conference on Information Systems (ECIS), pp. 1–16. Münster, Germany.

26.

Kimball

, & Ross

(2002). The Data Warehouse Toolkit: The Complete Guide to Dimensional Modeling, 2nd ed., New York, NY, USA: John Wiley & Sons, Inc.

27.

S.-M.

, Lu

, Tseng

K.-T.

, Chen

, & Chen

C.-L.

(2013). Energy-saving potential of the industrial sector of Taiwan. Renew Sustain Energy Rev, 21, 674–683.

28.

Min

, Shiwen

, & Yunhao

(2014). Big data: A survey. Mobile Network Applications, 19, 171–209.

29.

Müller

, & Loffler

(2009). Improving energy efficiency in manufacturing plants e case studies and guidelines. In: 16th CIRP International Conference on Life Cycle Engineering (LCE 2009). Cairo, Egypt, pp. 465–471.

30.

O’Driscoll

, & O’Donnell

G.E.

(2013). Industrial power and energy metering e a state of the art review. J Clean Prod, 41, 53–64.

31.

Philip Chen

C.L.

, & Zhang

C.-Y.

(2014). Data-intensive applications, challenges, techniques and technologies: A survey on big data. Inf Sci, 275(0), 314–347.

32.

Rohdin

, & Thollander

(2006). Barriers to and driving forces for energy efficiency in the non-energy intensive manufacturing industry in Sweden. Energy, 31, 1836–1844.

33.

Sarwar

, Persson

, Hellbergh

, & Haider

(2009). Measurement of specific cutting energy for evaluating the efficiency of bandsawing different workpiece materials. International Journal of Machine Tools and Manufacture, 49(12-13), 958–965.

34.

Seem

J.E.

(2007). Using intelligent data analysis to detect abnormal energy consumption in buildings. Energy and Buildings,, 39(2007), 52–58.

35.

Shrouf

, Ordieres

, & Miragliotta

(2014). Smart factories in industry 4.0: A review of the concept and of energy management approached in production based on the internet of things paradigm. Proceedings of the 2014 IEEE IEEM.

36.

Shrouf

, & Miragliotta

(2015). Energy management based on Internet of Things: Practices and framework for adoption in production management. Journal of Cleaner Production, 100(2015), 235–246.

37.

Srinivasan

, & Sheng

(1999). Feature-based process planning for environmentally conscious machining – part 1: Microplanning. Robotics and Computer-Integrated Manufacturing, 15(3), 257–270.

38.

Tao

, Zuo

, Xu

L.D.

, Lv

, & Zhang

(2014). Internet of things and BOM-based life cycle assessment of energy-saving and emission-reduction of products. IEEE Transactions on Industrial Informatics, 10(2), 1252–1261.

39.

Trianni

, Cagno

, Thollander

, & Backlund

(2013). Barriers to industrial energy efficiency in foundries: A European comparison. J Clean Prod, 40, 161–176.

40.

Tso

G.K.F.

, & Yau

K.K.W.

(2007). Predicting electricity energy consumption: A comparison of regression analysis, decision tree and neural Networks. Energy, 32, 1761–1768.

41.

Vijayaraghavan

, & Dornfeld

(2010). Automated energy monitoring of machine tools. CIRP Annals – Manufacturing Technology, 59(1), 21–24.

42.

Vikhorev

, Greenough

, & Brown

(2013). An advanced energy management framework to promote energy awareness. J Clean Prod, 43, 103–112.

43.

Waide

, & Brunner

C.U.

(2011). Energy-Efficiency Policy Opportunities for Electric Motor-Driven Systems. International Energy Agency, 2011.