Technological innovation in predictive maintenance

WTiN investigates technological innovation in predictive maintenance, and how it can ‘pay dividends’ and forge effective strategies within the textile manufacturing sector.

1. Introduction

New technology can deliver effective maintenance strategies to clothing and textile manufacturers, helping them go beyond reactive and proactive maintenance, moving into the more sophisticated world of prediction. The goal is to deliver an optimum maintenance strategy that enables manufacturers to get the most value out of their plant and equipment by spending the least amount of time, resources and money to deliver effective performance. Obviously, any machine benefiting from lavish daily maintenance will perform to the best of its ability, but some care could be a waste of money. How can companies best make tough decisions about when to invest in maintenance, of what kind and to what degree? That is what predictive maintenance is all about. Industry 4.0 machine learning (ML), artificial intelligence (AI) and sophisticated sensors are helping textile and clothing companies develop effective predictive maintenance systems. Researchers, vendors and consultants have been stressing the benefits, while warning about some potential pitfalls.

2. What is predictive maintenance?

Predictive maintenance is not new, even if tools and systems delivering this goal are being newly developed. Academics have defined predictive maintenance in many papers. A solid definition can be found in the Handbook of Natural Gas Transmission and Processing: “Predictive maintenance is [where] the service life of important parts is predicted based on inspection or diagnosis […] to use the parts to the limit of their service life. It manages trend values, by measuring and analysing data about deterioration and employs a surveillance system, designed to monitor conditions through an on-line system” [1].

Scientists note how additional savings over preventive maintenance can be achieved through portable instruments such as vibration monitors, thermography, ferrography, and more, to recognise impending machine failure. This paper, from US and Canadian academics, stresses: “A major benefit is the availability of an earlier warning, from a few hours to a few days, which reduces the number of breakdown ‘catastrophic’ failures” [1].

Predictive maintenance is usually conducted alongside preventive maintenance, of course, and involves detecting “small failures that begin the chain reaction that leads to big failures”.

A note from North Carolina, US-based asset maintenance solution provider Prometheus says that predictive maintenance is increasingly delivered through computerised maintenance management systems and databases connected through sensors to equipment [2].

By connecting sensors, factory or business-wide, a holistic view of maintenance status can be delivered. They can also deliver analysis to managers to plan maintenance strategies. And such systems enable companies to avoid running equipment until it fails or replacing while it still has operational life available – increasing costs and reducing plant efficiency.

Prometheus says: “Assume you have 50 technicians on your maintenance team and they each spend 20 hours a week performing routine inspections and preventive maintenance tasks – that’s 1,000 technician hours a week […] obsolete with a predictive maintenance strategy” [2].

3. Applications in clothing and textile production

From raw material assessment through to final product manufacturing, machinery in the textile industry can be fitted or retrofitted, with predictive maintenance programmes.

In general, according to John Alford, senior management consultant and maintenance subject matter expert with TBM Consulting Group, based in North Carolina, US, PdM is used to detect heat through thermography studies to identify ‘hot spots’ in equipment that signal potential problems. Most concerning are temperature spikes in electrical panels within drives, such as speed control systems, rotational systems in knitting machines and thread tension systems, which indicate a risk of a breakdown.

Alford explains how textile companies use infrared (IR) systems to detect hot spots. These can also be loose lug nuts (or bolts) which maybe generating friction, or a loose electrical connection which could cause a power system short or breaker to go down. Gearbox oil analysis and vibration monitoring are other key features that can spot early signs of potential defects, he says. Also, noise sensors, which monitor the background noise emitted by operating machinery – if that changes, it could indicate a machine is in poor health.

Another sensor that can be used assessed the number of ‘neps’ (entanglements in cotton fibre being ginned) which are supposed to be removed during carding, but blunted pins and wires can cause more neps, says research author Hitesh Manglani [3], a research assistant at the department of textile, chemistry & engineering (TECS) at North Carolina State University, US. If a sensor detects their numbers are increasing, it may indicate pins and wires need changing.

Other systems can detect changes to machine settings or components, such as multiple teeth being bent – all of which can alert operators to the need for predictive maintenance. A sensor used to check carding processes, for example, could be a temperature sensor that can avoid the cold starting of carding machines. It is important to ensure this equipment operates at optimum temperatures, as they can affect how their parts expand, affecting carding gap target settings. Other sensors can cull information on carded sliver and fluctuations in its weight from draw frames and comber machines, alerting control systems so they can change drafts to produce uniform sliver counts [3]. With draw frames, quality limits can be defined and predictive maintenance systems can shut down the draw frame if it exceeds certain tolerance limits, preventing damage. All analysed data can be generated in visual graphs, says Manglani [3].

Specialist system testers are constantly refining their PdM systems. Uster Technologies, based in Uster, Switzerland, prominent in the quality and testing instrument segment of the spinning industry, produces widely used cotton measuring and classification systems. Its Quality Expert product [4] gathers data throughout the manufacturing process and captures information on yarn irregularities and sources of defects along with density and variation, alerting staff to any anomalies indicating a potential maintenance issue.

Meanwhile, providing readily available data on yarn consumption, tension, and threading, monitoring systems developed by Vandewiele, from Kortrijk, Belgium, offers predictive maintenance [3]. Vandewiele offers digital models (twins) of both machines and production processes, that can be analysed for the optimisation of production settings, to improve scheduling, significantly reducing waste and generating savings in raw materials and energy use. Its technology is “paving the way to self-learning carpet weaving machines,” claims the company [5] and planning processes and make significant savings in raw materials and energy consumption.

Weaving machine suppliers Itema, based in Colzate, Italy, supplies LoomBrowser, management software to monitor machine efficiency which allows users to create or download new weft and dobby patterns [6]. It monitors the real-time health status of machine components, for example, tapes and sprockets, and suggests possible adjustments or interventions to allow machines to run safely at a higher speed [3].

Picanol, of Ieper, Belgium, has built in sensors to its machines to capture essential data and has managed to provide self-corrective functions to weaving machines [7]. This includes managing the lowest possible air consumption by controlling the opening and closing time of the relay nozzle valves according to the behaviour and air-transport-friendliness of filling yarn [3], reducing air consumption by 20% in some cases. It is also working on automating the removal of broken picks.

Finishing and technical textile machine producers Santex Rimar Group, from Vicenza, Italy, has developed an ATHENA software system using sensors to generate information on machine speed, internal machine temperature, machine vibrations, fabric tension, fabrics residual moisture quantity, and fabric width control [3]. It grants users access to real time planning, monitoring and control capabilities [8].

In the field of sewing, machine producers Juki, from Tokyo, Japan, provides JaNets (Juki Advanced Network Systems) to improve workforce efficiency and predictability. These are terminals at each workstation, monitoring and reporting on all production-floor activities, enabling management to identify when, where and why an operation stopped or was faulty. It can then eliminate the associated effects of a specific machine issue responsible for reoccurring errors to remedy larger production issues, for example amending the sewing speed if a large number of needles break on certain fabrics at a given speed, offering an early warning parts shortage alert and indicting the potential cost of a particular problem to the manufacturer [9].

In nonwoven production, the Osprey Corporation, an Atlanta, Georgia, US-based producer of air filtration and recycling systems, uses the OspreyCONNECT system [10] which includes a mentoring feature allowing HQ staff to see what a field technician is seeing on a piece of plant, sending documentation or information directly to managers via headsets or on computer screens.

4. Market perspective

How can predictive maintenance help save costs and disruption? According to industry digital transformation software specialists Pegasystems Inc, based in Massachusetts, US, “the size of the prize for predictive maintenance within manufacturing is immense”. The company adds that financially, there is a dire need to implement more effective maintenance practices in the manufacturing sector [11], the textile & clothing industry and others.

A report from Germany-owned Siemens Digital Industries Software [12] indicated that the cost of unplanned downtime is 10 times that of planned downtime. In its own briefing note, it quoted US Department of Energy statistics showing predictive maintenance can achieve 25%-30% return-on-investment (ROI) through lower maintenance costs. It can also decrease equipment failures by 70%-75% over the record of companies lacking such equipment and reduce equipment downtime by 35%-45%.

“Reactive maintenance costs four to five times as much simply because failed equipment reduces overall plant productivity, causes inventory backup and reduces overall efficiency,” it says [12].

“Companies eliminate the guesswork that characterises scheduled maintenance and instead can leverage insights based on real-time measurements, reducing and often eliminating errors” [12].

Such computerised maintenance management systems can also provide additional benefits, such as logging maintenance updates, delivering regulatory compliance data, assessments of labour costs, information aiding vendor management and produce data for budgeting.

Advantages such as these are a competitive benefit, and with sensors and production data becoming more widespread through the mainstreaming of Industrial Internet of Things (IIoT), the failure of companies to meet the new industrial maintenance benchmark of zero unplanned downtime will be noticed, the report warns [12]. “To remain competitive, companies must optimise asset performance and minimise failure, as even brief periods of downtime can result in reduced revenue, excessive overhead and strained resources,” says Siemens. “Margins are razor thin in manufacturing, so excessive maintenance costs and lost production time are unacceptable.”

Marzoli Machines Textile, based in Italy, is one vendor that stresses the real-world damage that can be caused by failures to predict breakdowns before they happen. Before a machine failure there is a sharp increase of energy consumption, temperature and vibrations with a reduction of yarn quality, the company stresses. These problems themselves are costly, but if detected early, they can be reversed through predictive maintenance. And, of course, if a machine is rendered unusable, there are many damaging problems caused [13]. Marzoli systems allow predictive maintenance data to be generated every 20 minutes, allowing real-time observations of machine health and indications about whether a component is going to fail.

Its remote maintenance system is available not only for all new Marzoli machines, but can be installed on older Marzoli and third parties’ textile production equipment through its retrofitting service which can be valuable in terms of delivering higher productivity and efficiency from older machines. This can also be cost effective in upgrading to a smarter factory without the expense of new machinery.

Moreover, Marzoli’s predictive maintenance systems use ML and Microsoft’s Azure Analytics to improve production efficiency by reducing downtime, shipping delays (by being able to predict failure and ordering spare parts before an existing part fails), and reduce spare part inventory, by buying in only what is needed [3].

Germany-based IoT and Industry 4.0 intelligence company IoT Analytics has reported that predictive general predictive maintenance initiatives saved organisations worldwide US$17bn in 2018 [14] driven by a surge in data points and sensor technology alongside ‘end-user skill shortage’.

It says the global annual market for predictive maintenance solutions is forecast to grow to US$23.5bn by 2024, by which time users will save US$188bn annually in costs, while benefiting from improved regulatory compliance and enhanced safety [14].

Pegasystems says those manufacturers deploying predictive maintenance reap benefits in relation to improved cycle times, quality, productivity, and customer experience, which in turn reduces costs and risks, thereby yielding tangible bottom-line results [11].

Systems can monitor and report on any actions required to correct a defect, with logs helping prevent them from recurring, meaning real-time adjustments can be made to processes to improve business practices and operational outcomes, reducing waste and lost time.

“The ability to troubleshoot and diagnose in real-time offers the manufacturer a competitive advantage in an industry challenged with low margins,” says the company [11].

5. Widespread use and future

While Pegasystems describes predictive maintenance as offering ‘a brave new world’ for manufacturers looking to improve productivity, underpin safety and deliver lower costs [11], it warns the introduction of such technology will take longer “and have more ups and downs, than many people might have expected”.

One problem is the plethora of small independent companies across the fragmented textile supply chain, who lack the financial muscle to invest in predictive maintenance technologies, relying instead on cheap labour, older equipment often ‘run to failure’ and a ‘low tech’ environment, TBM Consulting Group’s John Alford told says. While he adds, there are ‘world-class facilities out there’, running predictive maintenance systems, North Carolina State University’s Manglani concurs, saying IoT technology, on which predictive maintenance relies, is ‘not yet widespread’ [3]. He continues that a ‘clear financial and business appraisal’ of the benefits of IoT, is necessary to drive adoption.

His research indicates that a lack of skills and ‘poor interoperable standards’ across the supply chain could hinder progress in adoption of predictive maintenance. Some companies are averse to big data analysis, he says, suggesting that more supportive government policies and funding is required to help companies invest in relevant equipment.

However, he notes that across the supply chain, ‘every major OEM’ from processes including spinning, weaving, knitting, dyeing and finishing, sewing and nonwovens are moving towards IoT ‘in one form or the other’, persuaded to invest by the clear benefits of this technology.

Some equipment manufacturers, such as Vandewiele, are aware that offering modular systems that can be installed on a piecemeal basis can unlock investment. It allows mills to pay per package of predictive maintenance technology, so companies can assess the benefits and decide if they want to consider further investment.

And there are examples of companies installing such systems. A.Celli, an OEM in the nonwoven sector based in Lucca, Italy, formed a division called ‘extreme automation’ in 2016 to improve equipment efficiency via machine condition and energy monitoring and automatic defect detection mechanisms using artificial intelligence, for instance [15].

In its latest white paper [12], Siemens indicates that companies are seeking to benefit from the ‘sci-fi-sounding, near-magical benefits’ promised by IoT solutions, including predictive maintenance, as quickly and inexpensively as possible. And in his description of predictive maintenance [14], IoT Analytics managing director Knud Lasse Lueth refers to it as “the killer use case of the Internet of Things”.

But solution provider Senseye, from Southampton, UK, warns that ‘doing predictive maintenance well isn’t easy [16]’ and as implementing a PdM strategy affects all areas of a business, ‘it shouldn’t be taken lightly’.

Ultimately, it will be competitive advantage as well as cost-benefit that will drive this technology – and poorer, smaller companies with less capacity to manage high tech systems will adopt predictive maintenance more slowly than larger, higher tech rivals.

A good example of a company that has embraced such technology is India’s textile conglomerate Aditya Birla Group. It offers Birlasoft intelliAsset, a machine learning-backed predictive modelling solution that provides capabilities for asset visualisation, analytics and actions in near real time, the company says. “With intelliAsset, all your disparate data is on one system, visualised…in near real time, [offering data on] forecasting performance, predicting changes and prescribing action,” it says [17].

In Sri Lanka, the country’s leading apparel exporter Brandix says it uses predictive maintenance systems, such as infrared (IR) thermography inspections that help managers monitor the condition of electrical panel boards, identify potential risks and ensure proactive preventive action. Brandix’ fabric cutting machines are monitored by sensors that disseminate weekly reports enabling managers to check their condition.

While this is “not a real time monitoring system,” the company says, Brandix is “currently working on a predictive maintenance tool for facility-related machines [18].” Discussions are underway with ‘best-in-class service providers’ regarding predictive maintenance.

6. AI and ML

Without doubt, the increase in availability and sophistication of artificial intelligence (AI) and machine learning (ML) systems will make predictive maintenance more effective.  

AI consultant Alexandre Gonfalonieri, of Switzerland-based DNA Global Analytics, stresses how using AI and ML to draw insight from operation data ‘eliminates most of the guesswork’ from predictive maintenance decisions. Such automated systems utilise data from production sensors, operating computers (programmable logic controllers – PLCs), supervisory control and data acquisition (SCADA) data gatherers, enterprise resource planning (ERP) systems, manufacturing execution systems (MES) and more. AI and ML systems can help managers determine whether a failure is a one-off fixable problem, or an indication of terminal decline. It can warn what components are likely to be the cause of a failure. They can also be programmed to predict equipment’s remaining useful lifetime, and the risk of failure within a specific time-scan. This insight is useful [19].

However, to achieve these results, data fed into these systems must not only be accurate, but it must also be configured and integrated so that algorithmic analysis delivers useful rather than faulty results. Researchers are assessing which datasets maybe most useful in these systems. A December 2018 study from Morocco in the International Journal of Industrial and Systems Engineering [20] assessed the advantages and disadvantages of vibratory analysis; thermography; radiography; penetrant testing; magnetoscopy and ultrasonics, noting that their utility could vary depending on whether production under assessment was continuous, discontinuous, changed by project. The best dataset might also vary depending on the industry under predictive maintenance consideration too – for textiles it may be different from food production and automobile manufacture, for instance.

7. Sensor R&D

Patents are being developed to deliver data fuelling such sophisticated predictive maintenance within the textile sector. For example, Cristian Locatelli, of Brescia, Italy, filed for Italian digital solutions company Camozzi Digital [21], a US patent for a predictive maintenance monitoring system for spinning. This stresses that it is ‘extremely complex’ to implement an efficient predictive maintenance system, given that to secure accurate predictive data, integrated sensors need to be placed on a large number of parts within many spinning machines in a large number of spinning factories. The patent achieves this by linking a series of sensors to a “server, operationally connected to […] local transmission/reception means for global data storage,” delivering information for desired cloud-based statistical analysis needed to create effective predictive maintenance [22].

Another new US patent has been filed by Martin Gutberlet and Nitin T Patil for Swiss textile machinery manufacturer Maschinenfabrik Rieter. Here, the system involves reference values being created for configuring a network of sensors, with these statistical triggers stored in a database. Each sensor is read when a textile machine operates, with their signals matched to these stored reference values. “An alert or warning is generated to an operator of the textile machine when the correlated signal from one or more of the sensors shows a data irregularity,” says the patent [23].

These patents reflect how North Carolina’s Prometheus stresses that it is more complex to establish predictive maintenance than preventive maintenance systems, with appropriate infrastructure needed before strategies can be launched. “Collecting the right data to enable the organisation to accurately predict failures and malfunctions is crucial,” says Prometheus [2].

This must reflect the importance of assets, so that critical failures are avoided through triggering optimal maintenance of the right equipment at the right time. If it works, “these challenges can be easily overcome with a detailed plan of action”, the company states [2].

8. New predictive maintenance data and analysis systems

The textile sector does not want for predictive maintenance solutions offered by high tech vendors.

Equipment suppliers Valmet, from Espoo, Finland, is one. In January 2021, the company secured a contract to deliver machinery used in materials preparation, bleaching and drying to Stockholm, Sweden-based Renewcell at its new plant in Ortviken, which will upcycle cellulosic textile waste, such as cotton and viscose clothing [24]. Importantly, these machines will be fitted with predictive maintenance capabilities.

Switzerland’s Uster continues to innovate. Its new system Tester 6, introduced in 2019, works with its Quality Expert solution, gathering data from the manufacturing process and displaying information such as yarn irregularities and defects on dashboards and comparative reports for users. Tester 6 can measure sliver, roving, and yarn uniformity along with atmospheric test conditions [25]. Incorporating additional sensors can help manufacturers determine overall yarn hairiness, hairiness length distribution, yarn diameter, shape, density and variation, as well as yarn waste and dust.

Sensors inform users of any discrepancies in these areas which could pinpoint machine weaknesses. The instruments are all fitted with Uster's Knowledge Based System for detecting faulty machine elements via an extensive range of sensors.

Southern England’s Senseye has also unveiled a new product this year (2021), Senseye PdM Omniverse [16] - a knowledge platform helping industries plan and implement predictive maintenance operations. This is important, given a Senseye blog says analysts warn that many predictive maintenance projects fail, often due to a lack of monitoring expertise and available best practice to guide digital transformation projects [16].

In a company note, Robert Russell, CTO, says the new product was “a step change in approach to enabling predictive maintenance at scale and in a sustainable manner”, noting that knowledge required to implement effective predictive maintenance ‘is in short supply’.

But companies who do invest in such expert human resources can avail themselves of predictive maintenance digital twins – simulation-based, connected, virtual replicas of in-service physical assets, mirroring the life and experience of the asset [16]. Digital twins enable system the design and optimisation of predictive maintenance and optimise industrial asset management, being fed live data, mirroring the activity of physical asset.

Ansys, based in Pennsylvania, US, is one such company operating this technology, and indicates that digital twins, created with its Twin Builder solution, allow manufacturers to analyse predictive maintenance outcomes when planning installations, to save on warranty and insurance costs [26].

The Hugo Boss clothing plant in Izmir, Turkey, is one site where such technology has been implemented in recent years and it has paid dividends. The plant’s head of IT, Erkut Ekinci, told the 2019 IoT Solutions World Congress in Barcelona, Spain, that predictive maintenance allows managers to listen to the 2,500 different machines running in their factory. And if managers listen well, the machines ‘will say something about the future’ about maintenance, predicting failures, and utilisation rates [27].

Ansys agrees that the downstream fashion end of the business can also benefit significantly from predictive maintenance. It “should prove to be of tremendous value, given the fast-changing pace of fashion and consumption by end-product users,” the company concludes [26].

9. Vulnerabilities

In cybersecurity terms, there are different security requirements between information technology (IT) and operational technology (OT) and historically, “OT security is never done particularly well”, Gabriel Currie, from the UK threat detection and response team of accounting network PwC, says.

Such machines are tough to secure because they are typically old, with replacement cycles much longer than IT, because equipment is more expensive and harder to change: “Think about the challenge of swapping out components on a manufacturing floor vs your laptop,” he notes.

Another problem is that “there aren’t that many people who know how” to update electronic security on specialist operating systems used by one industry, says Currie.

Manglani warns there is a lack of built-in mechanisms to receive automatic firmware updates for OT systems and once installed the potential for upgrading IT security on these devices is often forgotten [3].

That means pre-existing security vulnerabilities, “are never or rarely updated,” he says, and these can consequently allow access by hackers to predictive maintenance systems and the data they hold.

Given production equipment is powerful, this “makes OT attacks quite appealing for attackers”. He adds: “If ransomware shuts down your manufacturing line it's arguably a lot worse than shutting down your marketing department”.

Predictive maintenance systems can also be uniquely exploited, says Currie, given their importance to predict/prevent the failure of parts – the very reason why this tech is being installed in the first place. Hackers accessing the OT network “therefore [have] an opportunity to compromise the integrity of the predictive maintenance system in order to hide failure until it’s too late,” Currie explains.

At particular risk, warns Manglani, are textile companies supplying items to the US Department of Defense [3], a target for foreign espionage and sabotage.

And as manufacturing analytics are cloud-based, this opens up a larger attack surface for would-be criminals, notes a paper from Israeli cyber-security experts Waterfall Security [28], as does the proliferation of sensors, upon which maintenance analytics rely [3].

10. Risks to the textile and apparel sectors

The risks of such attacks to operational technology incorporating predictive maintenance are severe. Threats can be categorised in three ways – harming availability of assets, loss of reputation and risks to confidentiality.

Disruption to the availability of materials, machines and labour leads caused by attacks on predictive maintenance systems can understandably cause significant productivity losses for the industry should the result be equipment being forced offline or being re-programmed to malfunction [3].

Raw materials sources, production data, customer lists, pricing strategies, future strategies and employee information can all be accessed if criminals attack the industrial control systems (ICS) or supervisory control and data acquisition (SCADA) systems – which are extremely vulnerable – via spoofing. Passwords and authentication processes are frequently lax in the industry, experts warn. [28]

This impacts the perceived integrity of a company should criminals attack card wire adjustments, and machine settings during a spinning stage, for example, causing defective products which, in turn, leads to reputational damage.

Attacks such as ‘denial of service’ hacks flooding an online system with traffic so that it breaks down can also cause disruption to supply chains, harming the availability of materials and causing downstream production disruption due to idling [3].

One company told Manglani in his research that “the difference in the mechanical parts of the machine will get smaller and smaller in the future, but the knowledge of how to combine and analyse data will increase. This non-functional difference will put the market in the focus of cybercriminals; in such a scenario, the data of the mill gets very important, and the loss of data can impact a company dramatically [3].”

Paul Bofin, senior business development manager with Fluke Reliability, based in Dublin, Ireland, says hackers can find key assets and bring them down [29].

“They can get into the asset management solution which says where something is, how old it is, what’s been done on it in the past, how it’s been maintained, they can read logs, identify problems and sabotage it,” he says. “They can take the logs and see, as a competitor, how inefficient it is.”

11. Protection

Cybersecurity professionals agree that the installation of correctly formatted firewalls, strong passwords and multi-factor authentication for user access are critical to securing preventive maintenance systems. A firewall with an intrusion detection system (IDS) is recommended by infrastructure monitors Ayyeka Technologies, based in Newark, New Jersey, US: “Humans cannot be programmed to detect when a suspicious person is looming on their property, but firewalls can,” says a company spokesman [30].

Manufacturers should also have strict protocols in place barring any employees or service contractors entering a production site from using USB memory devices in any part of the system, which can also open vulnerabilities.

“It is crucial to think like a hacker and identify the weakest link in the network and implement additional security measures,” says the Ayyeka spokesperson, recommending all communications between devices and the supervisory control and data acquisition (SCADA) systems, such as predictive maintenance networks, should be encrypted.

Companies should consider installing on-premises servers in addition to using cloud storage, their spokesman adds, with non-critical data available on the cloud and used as an auxiliary system, entirely outside a SCADA system.

Ideally, devices should not be connected via a fixed IP address and wait for incoming connections – “such devices are sitting ducks begging to be infiltrated” [30]. Instead, devices should use dynamic IP addresses and initiate data uploads to the network carrying data to an analytical hub only when needed.

A device’s communication with the control system should be limited to short intervals and should shut down once data transmission is complete thus restricting the hacker’s chance to locate the device, fraudulently complete authentication and decipher the encryption [30].

As data is gathered from sensors, it is also worth considering sensor fingerprinting, adding an additional layer of authentication when a device interacts with the network, meaning hackers cannot replace valid sensors with malicious devices [30].

12. Software and services

Security consultants are united in their belief that textile companies should continually test their connected products for potential security breaches. Fluke Reliability’s Bofin says that penetration-testing (pen-testing) is a crucial process. “You should do it – absolutely,” he says. Professionals can provide a service whereby they attempt to hack into a company’s system, identifying security loopholes and advising of additional safeguards to put in place.

Steve Zalewski, the deputy chief information security officer (CISO) for global apparel firm Levi Strauss & Co, says that he sees IoT pen-testing, including predictive maintenance, as a no-brainer: “These are untrusted unverified devices on my trusted networks,” Zalewski says. “This is immature technology without mature security controls” [31]. “We have got to bring some rigour” to IT security, he continues. “The problem with IoT devices is that security is not considered within the functional capabilities of the devices” [31].

In addition, Ayyeka emphasises that cybersecurity is ‘not a one-time investment’ [30] adding that every company needs to take into account every device that has communication capabilities.

“Cyber-protections require a holistic approach; a customised solution should be designed for each utility and operator” and involves assessing every component in the internal network and identifying the devices with communication capabilities supplied by external vendors.

Israel’s Waterfall [28] has developed a product, Unidirectional CloudConnect [32], which acts as a secure gateway between an industry’s site and the cloud. It collects data within a SCADA network using a wide variety of connectors, then publishes that data to the cloud using a selected safe protocol, protecting the company from remote online attacks.

13. Conclusion

Without question, the installation of predictive maintenance systems within the operating technologies of textile manufacturers is an investment worth making. The resulting extension of equipment life, improved productivity and avoidance of costly halts to production means there is a powerful case for investment.

And while many smaller textile firms might think this technology is beyond them (especially in low-cost outsourcing hubs), the truth is that sensible managers will always assess the potential of predictive maintenance. Companies offer modular installations that can be taken step-by-step. And while these systems require skilled managers and technicians to operate them, the value of recruiting quality personnel should always be considered.

Textile companies should take advantage of the opportunities offered by these systems to rethink how they operate. Can they be smarter in how they care for and maintain their productive equipment? They should be, because without looms, sewing machines, printers and knitting equipment, textile and clothing companies cannot operate. Machines are the partners of manufacturers, just like the human workers operating them. They deserve some care and attention too, and predictive maintenance systems help. Properly repaired machines always return the favour – with improved productivity.